Mapping scrub vegetation cover from photogrammetric point-clouds
Vafidis, Jim; Lucksted, Isaac; Gall, Moyrah (2022), Mapping scrub vegetation cover from photogrammetric point-clouds, Dryad, Dataset, https://doi.org/10.5061/dryad.0rxwdbs04
This dataset is derived from photogrammetric point cloud models of UAV imagery. It includes the Above ground models of vegetation as well as the isolated scrub vegetation.
We illustrate the method with two case studies from the UK. The scrub cover at Daneway Banks, a calcareous grassland site in Gloucestershire was calculated at 21.8% of the site. The scrub cover at Flat Holm Island, a maritime grassland in the Severn Estuary was calculated at 7%. This approach enabled the scrub layer to be readily measured and if required, modelled to provide a visual guide of what a projected management objective would look like. This approach provides a new tool in reserve management, enabling habitat management strategies to be informed, and progress towards objectives monitored.
Two readily available consumer-grade UAVs were used in this study. Daneway Banks was surveyed using a DJI T600 Inspire 1 quadcopter equipped with a 12MP Zenmuse X3 RGB sensor. Flat Holm was surveyed using the DJI T900 Inspire 2 with 20MP Zenmuse X4S RGB sensor. Both UAVs have comparable outputs, and choice of their use was driven by equipment availability and the greater stability of the Inspire 2 in more exposed conditions of the marine environment.
Transects were planned and conducted using Pix4D Capture (Pix4D China Technology Company) application running on a Sony Experia android smartphone. The transects were programmed using the ‘double grid for 3D model’ template, covering all areas of the site with front and side overlap setting of 80%. The cameras were angled at 70̊ (not nadir). The drone flying speed was set at ‘normal’ which was approximately 5m/s. The shutter speed, ISO and aperture for the both the Inspire 1 (Zenmuse X3) and Inspire 2 (Zenmuse X4S) was 1/2000, 200 and f2.8, with focus set at infinity. At Daneway Banks flights were undertaken on July 3rd 2017 at an elevation above ground level of 50m, giving a ground sample distance (GSD) of 2.49cm/px. The image capture at Daneway Banks collected 1127 images in five flights which took 78 minutes of flight time. At Flat Holm flights were conducted on 31st May 2019 at an elevation of 75m, giving a GSD of 2.34cm/px. The flight height was chosen on the basis of minimising disturbance to the colony of lesser black-backed gulls. The image capture at Flat Holm collected 1417 images in 7 flights, which took 106 minutes of flight time. Images were saved on SD-Cards as tagged image file format (tiff) including the GPS position, camera orientation, and time.
The data analysis workflow involved: Generation of a photogrammetric point cloud and associated elevation models, Generation of an Above Ground Model (AGM); Defining the study boundary, Classification of height bands, and; Measurement of the scrub layer.
Generation of a photogrammetric point cloud and associated elevation models; All UAV image files contain metadata of flight information (coordinates of UAV) and camera parameters (orientation, ISO, shutter speed and aperture). All images were uploaded to Pix4d Mapper V 4.5.6 which automatically produces geo-referenced orthomosaics and digital elevation models. Matching points are identified across all uploaded images and their 3D coordinates are calculated using Structure from Motion algorithms. The points are interpolated to form a triangulated irregular network, which generates a dense point cloud. This point cloud enables all image pixels to be positioned in the same scale on an ortho-rectified mosaic (or ‘orthomosaic’; Küng et al. 2012). In this study we use the ‘3D Maps’ standard template, which retains the full keypoints image scale in the initial processing. The point cloud densification was created at the original scale (1), at ‘optimal’ point density, and a minimum of three matches for each point. As well as the orthomosaic, Pix4d generates a Digital Surface Model (DSM) and a Digital Terrain Model (DTM) as exportable raster tiff files.
Generation of the Above Ground Model (AGM); To isolate vegetation from the ground and eliminate the effect from topographical variation, the DTM was subtracted from the DSM (DSM-DTM) using the raster calculator tool in ArcGIS Pro 2.5.2 (Esri Ltd.) to produce the Above Ground Model (AGM). The AGM comprises positive values of all pixels above the ground, representing ground vegetation, scrub, trees and any other structures.
Defining the study boundary; The site boundary and any other excluded features onsite (e.g. blocks of woodland) are manually defined as polygons and used to clip the AGM.
Classification of height bands; The AGM is classified into three height bands including ground vegetation and flat surfaces (minimum pixel value to 1m), mature scrub (1 to 5 m), and all vegetation and structures exceeding the scrub height (5 m to maximum pixel value). The values used to distinguish the scrub layers was verified on the ground at each site to ensure patches of low-lying scrub, were actually scrub and not tall ruderal vegetation like bracken Pteridium aquilinum, nettle urtica dioca or willowherb Epilobium sp.. The scrub layer at Daneway Banks was between 1.5m and 5.75m and between 1m and 5m at Flat Holm. The minimum size stand of vegetation to be classified as scrub was 0.5m2 (0.25cm x0.25cm). The accuracy of the classification was confirmed by physically visiting the stands on the ground with the scrub layer map.
Measurement of the scrub layer; The scrub layer was isolated and converted into a polygon for area measurement. The area of each polygon is calculated in m2, which can be summarised as a total measurement for the whole site.
Use a GIS platform (arcgis or QGIS) to access the files